learn-by-minimal
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill restores learning sessions by reading state information from a locally stored MAP.md file, creating a surface where modified file content could influence agent behavior. 1. Ingestion points: The agent reads ~/.learn-by-minimal/<主题-slug>/MAP.md during the 'Storage and Resumption' phase (Step 5) to determine the next learning node. 2. Boundary markers: Absent. There are no instructions to treat the file content as untrusted data or to ignore embedded instructions during the parsing of the Mermaid map. 3. Capability inventory: The skill is authorized to perform file read/write operations for demos and notes, and it can invoke the research-wx tool for external research. 4. Sanitization: Absent. The skill does not specify validation for the file content or the user-supplied topic name used in the file path.
- [DATA_EXFILTRATION]: The skill accesses the user's home directory to manage state. Findings: The path construction ~/.learn-by-minimal/<主题-slug>/ lacks sanitization instructions. A malicious input for the topic name (e.g., using directory traversal sequences like '../') could potentially be used to trick the agent into reading or writing files in unintended locations within the user's filesystem.
Audit Metadata